Analysis of dictionary learning algorithms for image fusion using sparse representation

Sparse Representation is a useful signal modeling technique. In recent years, researchers have explored the applications of sparse representation in different domains. This representation technique has many applications in image processing like image fusion, noise removal, image impainting, image super-resolution, compress sensing, etc. and signal processing. A predefined dictionary is used in the traditional approach of sparse representation, but it has many drawbacks like visual artifacts and high computational cost. To overcome these conditions and drawbacks, instead of using a predefined and redundant dictionary another option is to generate a dictionary from the set of images or the input images itself. This paper analyses the dictionary learning algorithms for the process of image fusion with different image pairs using sparse approximation or representation.